Dave runs a two-van electrical business. Last quarter he invoiced $84,000 across nineteen jobs and finished with less in the bank than the quarter before. His quoting is consistent, his hourly rate has not changed, and his materials supplier has not put prices up much.
He cannot tell you which jobs made money. Not because he is careless — his receipts are all there, scanned, categorised, totalled. They are just categorised as "Materials" and "Fuel", which tells him what he spent money on and nothing about what he spent it for.
This is the single most common gap in tradie bookkeeping, and the fix is one column.
The column
Add a job reference to every expense row. Not the client name, not the suburb — a short code that stays constant for the life of the job.
| Date | Supplier | Description | Amount | GST | Job | Category |
|---|---|---|---|---|---|---|
| 2026-07-14 | Middys | Cable, 2.5mm, 100m | 284.00 | 25.82 | J-0412 | Materials |
| 2026-07-14 | Caltex | Diesel | 96.40 | 8.76 | — | Vehicle |
| 2026-07-15 | Bunnings | Conduit, saddles, silicone | 118.60 | 10.78 | J-0412 | Materials |
| 2026-07-15 | Middys | Switchboard, 12-way | 640.00 | 58.18 | J-0415 | Materials |
Short codes beat descriptive names for the same reason short property IDs do: "Henderson kitchen" gets typed four different ways across a job, and then it does not group. J-0412 either matches or it does not.
Use data validation pointing at a job list tab so the code comes from a dropdown. A free-typed job reference is a job reference that will eventually be wrong.
The hard part: one receipt, several jobs
The template above is easy. The reality is the Bunnings run on a Tuesday morning where you buy conduit for one job, a box of screws that will last six months across everything, and a drill bit to replace the one you snapped yesterday.
One docket, three different kinds of cost. This is where most job costing quietly collapses, because the honest answer takes effort and the dishonest answer — assign the whole lot to whichever job you are driving to — is right there.
Three workable approaches, in order of how much effort they cost:
Split the row. Enter the docket as two or three rows, each with its own job code, totalling the docket amount. Most accurate, and for a $340 docket covering two jobs it takes thirty seconds with the itemised receipt in front of you.
Use a consumables job. Create a permanent code — J-STOCK or similar — for genuinely shared items: screws, tape, silicone, blades. These costs are real but not attributable, and forcing them onto a specific job distorts that job's margin. At quarter end you can spread the consumables total across jobs proportionally if you want a fully loaded view.
Assign to dominant job, and note it. For a docket where 90% is clearly for one job, assign the whole thing and put the exception in the notes. Less precise, but far better than not splitting at all, and it leaves a trail.
What matters is picking one rule and applying it consistently. Margins compared across jobs are only meaningful if the costs were allocated the same way each time.
Vehicle costs do not belong on a job
A diesel fill-up gets used across four jobs and three days of driving. Attaching it to a job because you filled up on the way to site is a fiction that makes the margin on that job look worse and every other job look better.
Leave vehicle running costs unassigned and treat them as an overhead against the whole business. If you want a per-job view that includes travel, a better approach is a notional travel charge — a fixed amount per site visit, applied consistently — rather than pretending the fuel was consumed by one job.
The same logic applies to insurance, tool purchases, phone, and licence renewals. They are the cost of being in business, not the cost of a job.
What the numbers actually show
Here is Dave's quarter once the job column exists, for five of the nineteen jobs:
| Job | Invoiced | Materials | Labour hours | Labour cost | Margin |
|---|---|---|---|---|---|
| J-0408 | 3,200 | 710 | 22 | 1,760 | 730 |
| J-0412 | 6,400 | 2,840 | 38 | 3,040 | 520 |
| J-0415 | 11,800 | 3,120 | 61 | 4,880 | 3,800 |
| J-0418 | 2,900 | 1,640 | 19 | 1,520 | -260 |
| J-0421 | 5,100 | 1,180 | 34 | 2,720 | 1,200 |
Two things jump out that no category-level report would have shown.
J-0418 lost money. Small job, materials ran to more than half the invoice — the kind of job that feels fine because it was quick, and quietly isn't.
J-0412 is the more interesting one. It is nearly twice the size of J-0408 and returned less margin. Same client type, same work, double the materials ratio. That is a quoting problem, not a cost problem, and it is invisible until you look per job.
Across nineteen jobs, a pattern of two or three J-0412s is exactly how a busy quarter ends with a disappointing bank balance.
The quoting feedback loop
Per-job costs are only half the value. The other half is comparing them against what you assumed when quoting.
Add two columns to your job list: quoted materials and actual materials. After a dozen jobs you will know your own bias. Most tradies underestimate materials on small jobs and overestimate on large ones, because large jobs get priced carefully and small ones get priced from memory.
That ratio is the most valuable number in the whole exercise. A sparky who learns that small jobs run 40% over on materials can adjust the quote, not the work.
Making it survive a real week
The job code has to be entered at capture, because by Friday nobody remembers which Bunnings run was for which site. The practical version is scanning the docket in the carpark before driving off, picking the job from a dropdown, and moving on — Scan2Sheet pulls the supplier, date, total, and GST off the image and lands the row with the docket photo attached, so the only thing left to do is pick the job.
Thirty seconds per docket, in the vehicle, while it is obvious. That is the whole discipline.
Starting without backfilling
Do not go back through the quarter. Start with tomorrow's first docket, run it for a month, and look at the numbers.
One month of accurate per-job costing across eight or ten jobs will tell you more about your business than a reconstructed year, and it is the version you will actually finish.